This proposal is a response to RFA-CA-92-21, Transfer of New Biostatistical Methods to Cancer Epidemiology. We will focus on the problem of measurement error of predictors, also known as errors-in- variables and including the problem of surrogate exposure information, although the related problem of error in the response will also be described. The area of measurement error modeling is one of the most vigorously developing areas in the theoretical biostatistical literature. However, for the most part there is little in the way of nontechnical discussions and reviews in the area, so that most applied biostatisticians are unaware of many of the most useful methods. In addition, there has been no attempt to provide good computer programs by implementing the techniques which are available. We propose to address the issue of transferring the literature in the area to applied biostatisticians and epidemiologists primarily by writing a nontechnical applied monograph reviewing the field. The monograph will include not only a survey of the literature in the field, but also formulae for computing standard errors and detailed analysis of specific data sets. After the monograph is written, we will ask the ASA to conduct a short course based on the monograph at its annual meeting. We will also address the issue of being able to compute the new methods appearing in the literature by developing high quality, fully documented computer programs written in C, which will allow ready availability on a variety of platforms. In addition to developing the programs themselves, we will allow them to be easily called in a natural language either on a PC or from a UNIX-based implementation of 5+. The code will be made available to all interested parties.
